In short
- A large multi-university study finds that faster AI means fewer people work.
- Economists now see real job losses along with significant economic growth.
- The debate has turned on whether AI will change the need for new jobs.
For years, financial experts were the experts who could tell you to ease any fears about technology. ATMs did not replace cashiers, Excel did not replace bookkeepers and robotic vacuums did not replace women. “Augment, not replacement” was the consensus.
Well, that consensus is falling apart.
A new paper from researchers at the Federal Reserve Bank of Chicago, the Forecasting Research Institute, Yale, Stanford, and the University of Pennsylvania surveyed 69 economists, 52 AI experts, and 38 superforecasters about how AI will reshape the US economy.
All three groups agree on one thing: Advances in AI mean fewer human workers. It’s a polite way to say “the few people who work.”
The numbers are staggering. Under what the researchers call a “fast” scenario – where AI surpasses human performance in many intellectual and physical tasks by 2030 – economists predict that the rate of unemployment in the US will drop from 62% to 54% by 2050.
About half of this decline, about 10 million jobs lost, can be attributed to AI rather than population or other factors.
Rapid events are not science fiction. It’s a world where AI can negotiate book deals, help in any factory or building, and replace all independent engineers, police officers, and customer service agents.
Anthropic CEO Dario Amodei he was forewarned that the crisis is more widespread than many expected—and the way the research is conducted is a good indication of its production. GDP tells the other half of the story.
Under these same scenarios, economists project annual GDP growth to hit 3.5% by 2045-2049—closer to post-WWII levels. AI experts are more optimistic, predicting growth of 5.3%. Creating more wealth, concentrated at the top, with fewer shareholders. The researchers show that under rapid AI, the richest 10% of households could own 80% of the total wealth by 2050—far more than pre-WWII inequality.
But there’s a nuance that’s often lost in the debate over AI jobs. The paper notes that the disagreement among experts is not only about whether powerful AI will arrive, but what happens to the economy when it does. That is a meaningful change. Previous pro-tech arguments assumed that even revolutionary machines could create new fields of work. A new question that economists are grappling with is whether AI, unlike ATMs, will create new jobs.
Meanwhile, the employment rate still appears to be stable. A Yale and Brookings studies as of the end of 2025 they have not found a sign of unemployment in almost three years since ChatGPT was launched. But research cited in the new paper a 13% drop in workers aged 22-25 for jobs that have been heavily exposed to AI. The macro is fixed. The edge is not.
In principle, economists and ordinary people are very different. Economists favor retraining programs (71.8% support) and strongly reject job guarantees (13.7%) and global start-up funds (37.4%). Ordinary people are more open to action. The authors of the paper say that the best policy depends on the situation – and right now, no one knows who it is.
So, the metaphor of “add, not replace” isn’t dead, but it’s on life support, and economists crunching the numbers have enough data to worry about.
Daily Debrief A letter
Start each day with top stories right here, including originals, podcasts, videos and more.
